Featured Hybrid Recommendation System Using Stochastic Gradient Descent

Abstract

Beside cold-start and sparsity, developing incremental algorithms emerge as interesting research to recommendation system in real-data environment. While hybrid system research is insufficient due to the complexity in combining various source of each single such as content-based or collaboration filtering, stochastic gradient descent exposes the limitations regarding optimal process in incremental learning. Stem from these disadvantages, this study adjusts a novel incremental algorithm using in featured hybrid system combing the feature of content-based method and the robustness of matrix factorization in collaboration filtering. To evaluate experiments, the authors simultaneously design an incremental evaluation approach for real data. With the hypothesis results, the study proves that the featured hybrid system is feasible to develop as the future direction research, and the proposed model achieve better results in both learning time and accuracy.